Abstract:
User reviews of mobile apps on platforms like Google Play and Apple App Stores are a rich and valuable source of
information for requirements engineering and software evolution. They reveal the users' needs, preferences, and opinions about the
apps and their features. However, extracting and classifying the non-functional requirements (NFRs) from these reviews is a
challenging task that requires sophisticated methods and techniques. In this research, we propose a novel approach that uses data
mining, natural language processing, and machine learning to automatically identify and prioritize the NFRs from user reviews of 99
top-rated games across four categories: Sport, Racing, Puzzle, Action and Casual. We collected 271,656 reviews from both platforms
and used feature extraction techniques to select and extract the most important NFRs from the reviews. We then used four machine
learning algorithms: Naïve Bayes, Support Vector Model (SVM), Decision Tree J48, and Logistic Regression (LR) to perform
sentiment analysis and rank the NFRs based on their importance and relevance. We focused on three types of NFRs: security,
flexibility, and maintainability. Our findings show that user reviews can help improve the outcomes of these NFRs and that our
approach can help developers understand their users and meet their needs from an NFR perspective, thus increasing user satisfaction
and retention.
